Section

Glossary

The reference layer. Alphabetical entries linking to the encyclopedia entries that ground each term.

33 terms

B

Blind Analysis: The practice of locking analysis methods before seeing the data, used in high-stakes empirical work to defend against bias toward expected answers.
Blind Replication Sprint: The hackathon archetype in which multiple independent teams work in isolation on a shared dataset using methodologically distinct approaches, with convergence of results as the success criterion rather than ranking against results.
Bounty: A sponsor-funded prize tied to narrow criteria specific to the sponsor's interests, central to the Sponsor-Bounty Federation format.

Calibration → Apples to apples: A pre-judging session in which judges grade a sample together to align their scoring.
Class A: A hackathon that has committed to a student fair-fight format. Eligibility is restricted to students or first-time hackers, pre-event project work is prohibited, and integrity mechanics apply to maintain the fair-fight commitment.
Class B: A hackathon that operates as an open competition. Funded teams are welcome, pre-existing work may or may not be allowed, and an optional separate prize pool for first-time hackers mitigates the dominance funded teams would otherwise have.
Class C: A hackathon that operates as a recruitment and pipeline event. Sponsors and recruiters are listed prominently; funded and pre-existing teams are not just permitted but expected — they are the point.
Code Sprint: A non-competitive concentrated-work event for an open-source codebase, distinguished from the competitive hackathon by the absence of rankings, prizes, and rubric-driven judging.
Convergence Event: The designed dramatic moment at which independent teams in a blind-replication protocol see each other's results for the first time and confirm or fail to confirm agreement.
Cross-Domain Bias: The documented unreliability of judge scoring when projects span non-commensurate problem domains.

Datathon: A Single-Problem Competition variant with a data-science audience, characterized by a single dataset, a single objective metric, and leaderboard-driven comparison.
Demo Bias: The documented tendency of judges to score teams with more polished presentations and more confident presenters higher than teams with comparable or stronger underlying work, particularly under science fair judging conditions.

Embargo: Enforced isolation between teams during a blind-replication protocol, designed to prevent cross-contamination of independent analyses.
Executive judging → Apples to apples: Final-round judging by senior executives, often less domain-specific than expert judging.

Finalist round → Apples to apples: The second-pass judging round that shortens the field after preliminary scoring.
Frame: The constraint that bounds what kinds of work belong in a hackathon. Themes are one kind of frame; civic problem areas, sponsor problem sets, and externally-provided frameworks like the UN Sustainable Development Goals are others.

Integrity Through Convergence: The integrity mechanic by which a hackathon's result is established through agreement across independent teams using methodologically distinct paths on a shared dataset, applied when no external ground truth is available on the day of judging.

Methodological Diversity: The deliberate use of methodologically distinct approaches by independent teams in a blind-replication protocol, ensuring that any agreement cannot be attributed to shared algorithm bias.

Outcome: A specific, falsifiable description of what success looks like at a hackathon. Outcomes pair with stakeholders to make a goal operational; together they are the difference between a designed event and a scheduled one.

Partner Prize: A sponsor-funded bounty in a Sponsor-Bounty Federation event, judged by the sponsor against narrow criteria specific to that bounty.
Pre-Event Work: Project development carried out before the official hackathon window opens. Class A events prohibit it; Class B events may permit it with disclosure; Class C events expect it.
Presentation: The structured demo segment in which a team surfaces its work to judges. NASA Space Apps weights presentation as one of five equally-rated criteria; the practitioner literature treats demo craft as part of the work rather than as an addition to it.
Problem Statement: The artifact that names what a hackathon is asking participants to solve, written in advance, owned by a real stakeholder, and pre-tested before publication. Stub.

Ringer Team: A team whose participation in a hackathon would be disqualifying at any reasonable Class A event but is not formally excluded by the event's published rules. Funded startups, hardware-rich teams, groups with pre-existing work, and teams with former-employee members of the sponsor are all canonical examples.
Rubric: The set of criteria against which submitted projects are evaluated. The choice of rubric architecture — single-problem, explicit tracks, or abstracted uniform — determines what apples-to-apples comparison the event is even capable of making.

Science Fair Judging: The judging architecture in which projects are arranged at tables in an exhibition hall and judges roam between them, scoring work through brief structured visits rather than through formal presentations.
Scope: The bounded space of work an event will accept as in-bounds, established by the event's frame and operationalized through problem statements and rules.
Stack ranking → Apples to apples: Ordering all entries head-to-head rather than scoring against a rubric.
Stakeholder: A real person or institution whose situation will be different after a hackathon ends — better if the work succeeds, unchanged if it doesn't. The stakeholder requirement is what makes a hackathon's goal operational rather than aspirational.

Theme: A particular kind of frame — a stated topical or domain constraint for a hackathon's work. Themes are valid frames when paired with stakeholder ownership and operationalized through problem statements; they fail when treated as substitutes for the goal-discipline they sit on top of.
Theme Track: A specific named track within a Themed Multi-Track event, characterized by a thematic constraint, a track-specific rubric, and a separate prize pool.
Track: A named subset of a hackathon with its own rubric and prize pool, used to keep judging apples-to-apples within a set of comparable projects. Tracks resolve cross-domain bias by refusing to make the comparison.

Vibe coding → AI era: Building software primarily by directing AI tools rather than writing code line by line. The Cognizant Guinness event made the term industry-legible.